New Versal AI Edge Adaptive SoCs Will Boost Scalar Compute

When Xilinx introduced the Versal AI Edge Series back in 2021, the company took the combination of AI compute capability and programmable logic it had put into its AI Core Series and cascaded it down to serve applications requiring lower cost and power. Now offered by AMD, the Versal AI Edge Series is adding a second generation of devices that will integrate more CPUs with a more efficient hardware accelerator to offer a single-chip embedded AI solution for automotive and IoT edge applications that need more scalar compute capability. But the parts won’t begin sampling until the first half of next year and won’t be in production until late in 2025.

The new Versal AI Edge Gen 2 parts won’t replace the first generation of devices, which were aimed mostly at edge sensing. Instead, they will target compute-intensive use cases such as automated driver assistance systems (ADAS) and video processing for smart cameras used in machine vision, healthcare, and industrial IoT. The devices are designed to meet automotive and embedded safety integrity levels specified by the International Standards Organization (ISO) and other safety and security standards.

Some models of the second-generation Versal AI Edge Series will have eight Cortex-A78AE and 10 Cortex-R52 CPUs, versus just two applications CPUs and two real-time CPUs in the first-generation devices. The new devices will also feature the second generation of the proprietary AI Engine ML, which AMD says will deliver a boost in AI inference efficiency and support for new data types.

AMD has not disclosed pricing information for the second-generation Versal AI Edge devices..